# US Stock-Bond Correlation and Capital Preservation > This report examines how the relationship between US stock and bond returns has changed across long-run annual data and a more recent monthly yield proxy. It considers what those measures can and cannot tell investors about preserving capital. Published: 2026-10-07 Publisher: BlackRidge (https://blckridge.com/) Canonical: https://blckridge.com/research/correlation-regime-shifts-20261006/ PDF: https://blckridge.com/research/correlation-regime-shifts-20261006/correlation-regime-shifts-20261006-en.pdf --- # US Stock-Bond Correlation and Capital Preservation This report examines how the relationship between US stock and bond returns has changed across long-run annual data and a more recent monthly yield proxy. It considers what those measures can and cannot tell investors about preserving capital. ## A shifting stock-bond correlation is no capital-preservation rule US stock-bond correlation changes sign across the three fixed samples. Every reported interval spans zero, so these estimates do not establish a persistent relationship [01] [02] [03] . - Period - Annual observations - Correlation - 95% interval - 1961–1980 - 0.08 - −0.65 / 0.50 - 1981–2000 - 0.37 - −0.03 / 0.75 - 2001–2020 - −0.51 - −0.81 / 0.06 ## US stock-bond correlation shifts across rolling windows The stock-bond relationship changed sign across the historical record. Its long-run pattern is less stable than a single hedge assumption suggests [01] [02] [03] . Nominal annual US stock and bond returns, 20-year rolling correlation coefficient, 1927 to 2020. The coefficient is dimensionless. The 75 windows overlap. Sources: [01] [02] [03] . - Measure - Value - Window - Minimum - −0.66 - 1998–2017 - Maximum - 0.52 - 1976–1995 - Positive - Negative ## A selected shift toward stronger negative stock-bond correlation The full annual scan selects 1998 as the split year. Stock-bond correlation turns negative afterward [01] [02] . Unitless full-scan Fisher-transform difference statistic across candidate annual split years, 1927 to 2020; 1,000 null replicates per stationary paired circular-block setting [01] [02] [03] [05] [06] . - Block length (years) - Adjusted p-value - 95% maximum critical value - 0.038 - 3.13 - 0.090 - 3.54 - 0.173 - 4.09 ## When stock-bond correlation turns positive Across annual observations from 1927 through 2020, stock-bond correlation is negative below 4% average CPI inflation and positive at or above it. This retrospective split has only 25 higher-inflation observations. - Group - n - Correlation - Nominal joint loss - Inflation-adjusted joint loss - High inflation - 0.29 - 4/25 (16.0%) - 9/25 (36.0%) - Low inflation - −0.09 - 3/69 (4.3%) - 4/69 (5.8%) - Threshold - High n - High correlation - Low n - Low correlation - 3% - 0.14 - −0.05 - 4% - 0.29 - −0.09 - 5% - 0.28 - −0.08 ## When stocks and bonds fall together Joint losses were uncommon in nominal terms and more frequent after annual inflation adjustment. That distinction matters for capital preservation, although sign counts hide loss size and the path within each year. ### Nominal returns ### Returns after annual CPI adjustment - Measure - Count or rate - Denominator or interval - Bonds positive when stocks lose - 16/23 (69.6%) - Bonds exactly flat when stocks lose - Real joint losses - 13.8% - 7.4% / 21.3% - Independence reference, real losses - 13.2% - Excess over independence, percentage points - 0.6 Percentage points - −4.95 / 5.89 Percentage points ## P60 drawdowns: nominal and real recoveries diverge A portfolio can regain its nominal high while still buying less than before. P60’s two drawdown paths show how inflation changes the timing and depth of capital recovery. Annual percentage decline from each line’s own previous peak for a 60% US stock, 40% bond portfolio rebalanced annually, shown nominally and after annual-average CPI adjustment, 1927-2020. [01] [02] [03] [07] - Series - Drawdown - Peak - Trough - Recovery - Duration (years) - Nominal - −44.0% - 1928 - 1931 - 1935 - Real - −37.9% - 1972 - 1974 - 1984 ## US stocks and bonds: a stock/bond price-direction proxy This stock/bond price-direction proxy pairs US broad-market monthly stock total returns from the Kenneth R. French Data Library with the negative change in the last valid business-day 10-year Treasury yield, in percentage points. It is not a bond total-return correlation. [01] [04] [06] Monthly stock/bond price-direction proxy: rolling 36-month correlation of US broad-market monthly stock total returns and the negative change in the last valid business-day 10-year Treasury yield, with yield changes in percentage points. January 2016-August 2026, latest endpoint August 2026. The plotted correlation is unitless. Kenneth R. French Data Library vintage 202608, FRED yield observations through October 2, 2026. [01] [04] [06] - Period - n - Proxy correlation - Joint direction count - 2016-01 / 2020-12 - −0.40 - 2021-01 / 2025-12 - 0.54 - 2026-01 / 2026-08 - 0.15 - 2022 measurement - Value - Unit - US stocks, 2022 total return - −19.9 - % - 10-year Treasury yield change, 2022 - 2.36 - percentage points ## Stock-bond correlation changes modeled portfolio risk The weights stay fixed. Only the assumed correlation changes in this hypothetical stress, while annual stock and bond standard deviations remain constant. - Correlation scenario - Annual volatility, % - −0.5 - 9.6% - 0.0 - 11.9% - 0.5 - 13.8% - 1.0 - 15.4% ## When low volatility still loses purchasing power Capital preservation depends on the loss path. From 1927 through 2020, the 60% equity, 40% bond control portfolio recorded a 44.0% nominal maximum drawdown and a 37.9% inflation-adjusted maximum drawdown. [01] [02] [03] Annual inflation-adjusted drawdown from prior peak, percent, 1927-2020. [01] [02] [03] - Allocation - Tracking error - Nominal maximum drawdown - Inflation-adjusted maximum drawdown - Inflation-adjusted CAGR - P60 - 0.0% - −44.0% - −37.9% - 5.8% - P40 - 4.4% - −30.5% - −31.9% - 4.8% - P80 - 4.4% - −57.3% - −50.2% - 6.5% - 100% bills - 12.9% - −0.1% - −46.3% - 0.5% ## US Stock-Bond Correlation and the Limits of Capital Preservation Across the full record, the stock-heavy mix compounded at 5.8% a year after inflation. Paths diverged. [01] [02] [03] Index points, cumulative deflated wealth with end-1926 set to 100; annual observations from 1927 through 2020; [01] [02] [03] . - Allocation - Nominal CAGR - Deflated CAGR - Worst deflated year - Deflated drawdown - Real loss years (of 94) - P60 (60/40/0) - 8.7% - 5.8% - −24.0% - −37.9% - C25 (45/30/25) - 7.5% - 4.6% - −18.7% - −29.7% - C50 (30/20/50) - 6.2% - 3.3% - −13.4% - −24.7% - Period - P60 deflated CAGR - C25 deflated CAGR - C50 deflated CAGR - 1961-1980 - 1.6% - 1.4% - 1.2% - 1981-2000 - 9.8% - 8.2% - 6.6% - 2001-2020 - 5.8% - 4.3% - 2.7% ## Longer horizons saw fewer purchasing-power losses P60 is a 60% US stocks, 40% bonds portfolio, rebalanced annually, with returns adjusted by annual-average CPI. Longer windows ended below starting purchasing power less often. No guarantee follows. US CRSP P60: 60% stocks, 40% bonds, annual rebalancing. Annual-average-CPI-adjusted terminal wealth index per 100 at window start. 90 overlapping five-year windows, 1931 to 2020. [01] [02] [03] - Horizon - Windows - Loss windows - Loss frequency - Minimum terminal wealth - 30.9% - 76.0 - 12.2% - 75.7 - 8.2% - 76.3 - 0.0% - 106.0 Non-overlapping five-year windows anchored at 1927: 4/18, 85.7 ## Sources and measurement boundaries The annual record covers 94 years, 1927 to 2020 [01] [02] [03] [07] . A separate monthly diagnostic covers 128 paired observations, January 2016 to August 2026 [01] [04] , with no splice. Returns are in USD, gross of fees and taxes, and describe historical indexes rather than investable products. Monthly gross US stock returns are formed by adding Mkt-RF and RF within each month, then compounding 1+(Mkt-RF+RF)/100. Bills are compounded separately as 1+RF/100. Vintage 202608 was downloaded on 6 October 2026. [01] Release 6 provides nominal reconstructed US bond returns and headline annual-average CPI. The paired annual sample contains 94 complete years from 1927 through 2020, with no year excluded for missing input. [02] RORE documentation describes original US bond instruments and splice choices. Returns are yield-imputed for 1927-1928, then use long government bonds from 1929 through 2015 in the historical series. [03] The original Federal Reserve H.15 release reports the 10-year Treasury constant maturity yield in percent. FRED distributes this series as DGS10. It measures a yield, not a bond total return. [04] The December 2023 BIS Quarterly Review examines mechanisms behind stock-bond comovement. The research describes observed association and does not establish that those mechanisms caused comovement in any particular sampled period. [05] Recent macroeconomic events caused short-lived changes in bond-stock comovements, the authors report. High risk premia tend to coincide with large absolute comovements. These findings qualify claims of permanence. [06] Table 2 documents US 2016-2020 Datastream 10-year government bond index total returns, while its CPI table covers RENT price indices. Headline annual-average US CPI comes from the linked database [02] and JST release 6. [07] ## Keep up with the research. Choose optional research updates, or learn how the account structure works. ## Independent research, introductions clearly defined. BlackRidge is an independent research bureau introducing private investors to quantitative traders through multiple strategy providers. We publish quantitative research. Investors access strategies through PAMM accounts at the broker. We are not a fund or broker and never hold client money. Your funds stay at the broker. The account is opened in your own name. BlackRidge never receives, holds, or has withdrawal rights over client capital. No fee on your profits. A one-time access payment of 6.7% of the agreed trading level, paid to the selected strategy provider. No management fee, no performance fee, no profit share, in any year. You fund the risk, not the exposure. Allocations are notionally funded: you agree a trading level and fund the margin and drawdown allowance behind it. Trading losses can exceed the deposit without applicable negative balance protection; the separate 6.7% access payment is nonrefundable. blckridge.com/research : published research. Notional funding and PAMM accounts explained in full, and answers to the questions this raises. We use AI models to gather and aggregate source material and to help prepare each report. Read it with its cited sources, sample, methods and limitations, and send corrections through research methodology and corrections . The client pays the selected strategy provider a one-time access payment of 6.7% of the agreed trading level. BlackRidge receives an introduction fee from that provider and does not receive broker compensation. About BlackRidge . Minimum funded capital $25,000. The strategy, its live records, the broker, the selected strategy provider and the full commercial terms are presented on an introductory call and confirmed in writing before any payment or deposit. This report is published for information only. It is not investment advice and does not take account of your circumstances. Trading leveraged instruments including CFDs carries substantial risk and is not suitable for all investors; you may lose the capital you fund and, depending on your broker's terms, may owe more than your deposit. Findings in this report may combine third-party evidence, historical calculations and illustrative scenarios; they are not verified live trading results of any strategy provider. Past performance does not indicate future results. ## Sources - [01] [Kenneth R. French Data Library, US broad market stock series](https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/f-f_factors.html). Also: [[01]](https://blckridge.com/research/correlation-regime-shifts-20261006/#src-01). Monthly gross US stock returns are formed by adding Mkt-RF and RF within each month, then compounding 1+(Mkt-RF+RF)/100. Bills are compounded separately as 1+RF/100. Vintage 202608 was downloaded on 6 October 2026. [01] - [02] [Jordà-Schularick-Taylor Macrohistory Database, release 6](https://www.macrohistory.net/database/). Also: [[02]](https://blckridge.com/research/correlation-regime-shifts-20261006/#src-02). Release 6 provides nominal reconstructed US bond returns and headline annual-average CPI. The paired annual sample contains 94 complete years from 1927 through 2020, with no year excluded for missing input. [02] - [03] [RORE documentation, original US bond instruments and splices](https://www.macrohistory.net/app/download/9834516469/RORE_documentation.pdf). Also: [[03]](https://blckridge.com/research/correlation-regime-shifts-20261006/#src-03). RORE documentation describes original US bond instruments and splice choices. Returns are yield-imputed for 1927-1928, then use long government bonds from 1929 through 2015 in the historical series. [03] - [04] [Federal Reserve H.15 release, 10-year Treasury constant maturity yield](https://www.federalreserve.gov/releases/h15/). Also: [[04]](https://blckridge.com/research/correlation-regime-shifts-20261006/#src-04). The original Federal Reserve H.15 release reports the 10-year Treasury constant maturity yield in percent. FRED distributes this series as DGS10. It measures a yield, not a bond total return. [04] - [05] [Bank for International Settlements, Quarterly Review, December 2023](https://www.bis.org/publ/qtrpdf/r_qt2312v.htm). Also: [[05]](https://blckridge.com/research/correlation-regime-shifts-20261006/#src-05). The December 2023 BIS Quarterly Review examines mechanisms behind stock-bond comovement. The research describes observed association and does not establish that those mechanisms caused comovement in any particular sampled period. [05] - [06] [Campbell, Pflueger and Viceira, “Bond-Stock Comovements,” NBER Working Paper 34323, revised June 2026, Annual Review of Financial Economics 2026](https://www.nber.org/papers/w34323). Also: [[06]](https://blckridge.com/research/correlation-regime-shifts-20261006/#src-06). Recent macroeconomic events caused short-lived changes in bond-stock comovements, the authors report. High risk premia tend to coincide with large absolute comovements. These findings qualify claims of permanence. [06] - [07] [JST release 6 returns-extension documentation](https://www.macrohistory.net/app/download/9918957869/JST_RORE_Documentation_R6.pdf). Also: [[02]](https://blckridge.com/research/correlation-regime-shifts-20261006/#src-02), [[07]](https://blckridge.com/research/correlation-regime-shifts-20261006/#src-07). Table 2 documents US 2016-2020 Datastream 10-year government bond index total returns, while its CPI table covers RENT price indices. Headline annual-average US CPI comes from the linked database [02] and JST release 6. [07] --- ## Citation context US stock-bond correlation changes sign across the three fixed samples. Every reported interval spans zero, so these estimates do not establish a persistent relationship [01][02][03]. The annual record covers 94 years, 1927 to 2020 [01][02][03][07]. A separate monthly diagnostic covers 128 paired observations, January 2016 to August 2026 [01][04], with no splice. Returns are in USD, gross of fees and taxes, and describe historical indexes rather than investable products. Read signs as sample-specific. Correlation alone does not quantify portfolio losses or purchasing power after inflation. This boundary matters. No standing portfolio rule follows. Evidence stays limited. Primary input: [Kenneth R. French Data Library, US broad market stock series](https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/f-f_factors.html). Stable permalink: [https://blckridge.com/research/correlation-regime-shifts-20261006/#citation-context](https://blckridge.com/research/correlation-regime-shifts-20261006/#citation-context). ## Read next - [Historical Covariance Forecasts Underestimate Stress Risk by a Factor of 1.6](/research/covariance-estimation-error-20260929/). The related study tests how historical covariance forecasts behave during market stress.